Pressure Advance tuning issue: gap at corner

[blalor]

I’m new to Klipper and I’ve just built a Voron V0(.0), my first DIY printer. Nothing like stepping off into the deep end. I’m attempting to tune Pressure Advance on this printer. I’ve already done Input Shaper with an ADXL345; I’ve ended up with

[printer]
max_velocity: 300
max_accel: 9000

[input_shaper]
shaper_type_x = zv
shaper_freq_x = 104.2
shaper_type_y = zv
shaper_freq_y = 119.6

I’m tuning Pressure Advance with this startup g-code, per the guide:

SET_VELOCITY_LIMIT SQUARE_CORNER_VELOCITY=1 ACCEL=500
TUNING_TOWER COMMAND=SET_PRESSURE_ADVANCE \
    PARAMETER=ADVANCE \
    START=ss \
    FACTOR=ff

The image below details my results thus far.

I have additional photos I’ll attempt to link below, but I’m too new to the forum to embed them in this post.

My first attempt, using START=0 and FACTOR=0.020 (I’m using a bowden setup) and the slicer-default retraction of 2mm seems to have progressed too quickly past the sweet spot and left a massive gap at the left, after the corner is turned (the nozzle printed from left to right on the faces showing in the photo).

For my 2nd attempt, I realized the PA was advancing too quickly, so I tried using the direct-drive FACTOR of 0.005; the corner gap didn’t appear, but at the max PA value of 0.25 for this print, the corner still wasn’t sharp.

My 3rd attempt started where the 2nd left off, START=0.25, but with FACTOR=0.007 to speed up the progression a bit. I think the best looking corner is about ¼ of the way up, there’s still a valley on the left where the nozzle starts its acceleration.

For kicks, I lowered the retraction to 0 and repeated the previous attempt, START=0.25 FACTOR=0.007, and I don’t see a huge difference.

The problem I see with all of these is that they don’t seem to match the reference photo in the guide. I would have expected a properly-tuned PA to have sharp corners at both the accelerating and decelerating sides. Am I over-thinking this?

Thanks!
Brian

[blalor]

First attempt, START=0 FACTOR=0.020, retraction 2mm:

[blalor]

Second attempt, START=0 FACTOR=0.005, retraction 2mm:

[blalor]

Third attempt, START=0.25 FACTOR=0.007, retraction 2mm:

[blalor]

Final attempt, START=0.25 FACTOR=0.007, zero retraction:

[blalor]

Another angle:

[blalor]

(phew, last one!)

Finally, my printer.cfg.

[Sineos]

On these PA tuning tests you will always get one really bad side where the seam is.
Judge the effect on the remaining 3 corners to choose the proper PA value.

My personal advice is to rather choose the lowest good value instead of the highest good value.

[blalor]

All 4 corners are doing the same thing. After turning the corner, it’s underextruding, but looks normal as it approaches the next corner.

I will definitely go for less PA.

[koconnor]

In general, I look for the nicest looking corner. Sometimes it’s easy (as in some of the online guides) and sometimes it’s hard to pick a spot.

If you feel your corners aren’t good at any level, then it might be the extruder settings. For example, you may need to tune the temperature or reduce the flow rate (eg, run the test at a lower speed than 100mm/s or with a smaller layer height). It may also be a tricky filament (for example, PETG can be difficult to extrude - it often needs lower speeds).

Finally, using zv for input_shaper is a bit unusual - normally mzv is preferred. It may be fine, but if you’re unsure you could try disabling input_shaper for a test (just to verify it isn’t introducing poor corners).

-Kevin

[blalor]

Thanks @koconnor; I expected to see rounded corners as the PA increased, not an increasing gap with no/underextruded filament. As a rank amateur here, it’d be useful to me to see the PA guide expanded a bit with examples of what I might see. I also found that the “bowden” factor of 0.020 was far too high for my printer (which has a bowden tube length of about 41cm). It’d also be helpful to mention that the factor may need tweaking based on what your results look like. I’ll put together a PR with some documentation updates based on my experience.

using zv for input_shaper is a bit unusual

I’ve read that, but decided to not second-guess the machine and took the results of SHAPER_CALIBRATE verbatim. I had similar PA calibration results without input_shaper enabled at all. I’m still getting my rough tuning done (I’ll definitely need more retraction than 0.75mm with PA enabled), but the results are promising. My benchies already look better than what I get with my Prusa MK3S and Octoprint!

[Sineos]

@blalor

I also found that the “bowden” factor of 0.020 was far too high for my printer (which has a bowden tube length of about 41cm)

This is actually quite surprising. Before moving to a direct extruder I had about the same length of bowden. I went through quite some iterations in improvements and finally ended up with original Capricorn bowden and hydraulic screw fittings (instead of pneumatic push fittings).

All the measures brought down the needed PA considerably but nevertheless I always ended up between 0.4 to 0.7 depending on material. Something which is also true for people around me who also use Klipper.
For this range the factor of 0.02 fits really well.

I never experienced such a gap with the materials I print (PLA, PETG, ASA, ABS, PC) neither with my bowden, nor with the direct feed. Also not with 0.4 nozzle or 0.6 nozzle.

@blalor

I’ll definitely need more retraction than 0.75mm with PA enabled

Agreed. Even with the direct extruder I settle around 1 to 1.5 to avoid stringing

[Foko]

@blalor, @koconnor, i have the same issue with my bmg clone and v6 clone with a ~350mm bowden… . I get a gap (looks like underextrusion - from 0 to 50mm) after the printer moves out of the corner.

@blalor do you solve the problem?

Thank you!

[blalor]

@Foko it ultimately comes down to picking the height at the corner that looks the best (or averaging the ones that look the least bad). The slicer settings for this test are intended to exaggerate the effects of PA. I don’t think you should be getting a gap with zero filament, but there may be one side of the corner that’s not flush with the plane on that side. Make sure the extruder’s not skipping steps; I’ve had PETG filament that I cannot get a good PA tune for without it doing that.